Can I lower volume output with any command in lame enc 3.97b3 or maybe with RazorLame?

I rip CDs with EAC and later convert the wav to mp3 with latest lame enc 3.97b3/ RazorLame latest version.
Mostly I keep both lossless wav and the lossy mp3 for my mp3 player
When converting the mp3 file back to wav to see if it created any vol clips in Sound Forge, I allmost always get volume clips when using clip detection.

Don't want to normalize, only lower the vol!
Easy done with Volume in SoundForge but I do not want to rip the CDs with SoundForge coz it isn't secure as EAC.

Know that EAC and dbPowerAmp has this but I rather not use it coz I always get better results with RazorLame.

I just need to lower the volume on the final mp3 by 0.4 - 0.6dB / about 94 - 95%, I do not want to change the original wav files.


Hope someone can tell me whats the easiest way to do this process.

You can use mp3gain for that. Volume can be changed in 1.5 dB increments, IIRC, and it gives you indication about clipping or not. To do what you want you can use its "Apply Constant Gain" function, but you can also take into consideration the possibility to normalize your albums at "standard" average level of 89 dB using the "Apply Album Gain" function.

BTW, volume changes made by mp3gain are "lossless", that is it doesn't imply a transcoding. It just fiddle with some parameter into the mp3 frames and the changes you make can be reverted.

Cheers.

Sergio

I'd just use foobar2000 to decode MP3s, with RG applied, no preamp, and clipping prevention on.

PS. If you don't want RG attenuation, choose source mode: none, and processing: prevent clipping according to peak. You'd still have to scan the files for RG info, though.

tried that one.
But I need to change the wav file before converting to mp3.

I see no point in changing the final mp3 file,


If you use mp3gain, then you still have the clip but on a lower volume output.
The wave is still cut, the more you keep under that limit the more of the original sound you can save.


about clipping prevention in foobar:
this changes the file, limits the output.



to put it simple, I want to keep as much of the original structure as I can

I process my MP3 albums (which I convert from my lossless archive) with MP3gain's "Maximum no-clip album gain" option. It keeps the relative track volume levels inside each album correct and prevents the tracks from clipping with all decoders (HW or SW).

With modern compressed pop/rock music it mostly changes the overall album level -1.5 dB. Sometimes -3.0 dB or even -4.5 dB is needed. I don't use the "undo tags" option since there's no reason to make the files to clip again. After that I analyze the files with foobar for creating the standard replay gain tags so I can choose between the album and track gain modes on PC playback.

In your first post you said that:


This can be a sign of clipping due to the mp3 decoding process. The solution I suggest will solve this issue.

If instead the clipping is already in the original wav, as it was ripped from the CD, then blame the mastering engineer and... there is nothing to do. Nobody will give you back information that is not available in the first place. You can lower the wav volume using WaveGain (available at http://www.rarewares.org/others.html), but that yes, will just lower the level of the clipped parts without "regenerating" them (an impossible process from a deterministic point of view, at most some program can exist to extrapolate interpolate the missing values).

Sergio

I know, but as I said before, I rip the CDs with EAC to wav.
Ofcourse I have checked the original file.

The issue is, when encoding, clipping will be a fact if the original wav file is on max vol output/ 100%.

try and you will see

the sound wave changes and that change creates the clipping.
there is no point of removing clips if the clips are there.

I want to prevent clips.

I can do this with SoundForge I just need to make this process faster or automatic somehow

MP3 files have correct unclipped waveform before decoding even the standard decoding process makes clipped 16-bit pcm output. MP3Gain (or e.g. foobar2000's replay gain) can change the volume level before decoding to 16-bit integer so that the decoded files remain unclipped (only the volume level is a bit lower).

I perfectly know what you mean, believe me, man. The problem is in the decoding part of the process. You try, you then tell!

Cheers!

Sergio

Edit: BTW, if you don't trust me, just try to search the forum (always a good idea). This issues has been already debated several times and the solution already offered several times.

Two weeks ago I posted a screenshot that illustrates how replay gain correction can prevent MP3 files from clipping: http://www.hydrogenaudio.org/forums/index....st&p=421963

ok :S
The sound wave gets destroyed when you increase the vol "over the top" on a original wav file.
You can't get that back as it was when lowering back the volume.
So how am I gonna do it on a mp3 file if I can't on a wav file

As I said, I want to prevent clipping not edit the mp3 file in anyway, even if I do not need to re-encode it

I just need to make the process faster or automatic somehow.

When ripping a CD, I do not want to touch the final mp3 file, just play it

The CDs either have clipping or they don't. You can't do anything about that.

The MP3s either have clipping or they don't. You can't do anything if it inherited clipping from the CD.

You can do something if there wasn't any clipping on the CD. You can turn down the volume at decode-time (or use MP3Gain) and you'll get unclipped output.

That's what people are trying to tell you.

I'm tempted to give up...

case of where_is_it_clipped:

A) in the original, that is on the pressed CD. Do you think that can be a solution for that? NO!

B) in the process of encoding/decoding to/from mp3. By the heck, it is in the D_E_C_O_D_I_N_G (can you hear me?) part of the process an NOT (by definition) in the encoding.

mp3gain alters the "global gain" field of each mp3 frame so that it doesn't clip any more when (you guess when?) D_E_C_O_D_I_N_G.

Sergio

Adding --scale .95 to the LAME command-line parameters should do the trick.

No, just turning the volume down will not prevent clipping, it will just lower the volume of the clippings.

What moneo suggested (to use --scale 0.95) can do the trick, but you can't be sure 100% and besides that it alters the original (as probably SoundForge does) and personally I don't like that. I prefer lossless alterations, and for a reason, I think.

Sergio

OK, I should have said "You can turn down the volume using RG...".


On second thought, I don't think that's true. The MP3 format is in floating point. Keep lowering the volume and the peaks will come down to less than the maximum integer peak.

Oh, yes! That yes! Sorry for the misunderstanding. In fact it is what I do. Just store ReplayGain information in the ID3v2 tags (not even alter the "global gain field") and use foobar or winamp with otachan plugin for listening without clipping. Anyway recently I bought a DAP and I'm now tempted to move to the "global gain" field solution, so that even my DAP doesn't clip...

Cheers!

Sergio

What Moneo suggested works if the original wav file has no clipping.
Thanx Moneo !!!
smz, I know what you mean but the original wav I have tried with has no clipping and I do not want to create clipping in the mp3 file of a wav that does no have clipping


When playing a mp3 it gets decoded while playing.

If I encode a orignal wav file (that doesn NOT have clipping) to a mp3 file (256) and decode that mp3 to a new wav file with the same lame.
How can the mp3 have no clipping and the decoded file have when in the end of the play or decode process you will have the same audio

It depends on where the volume adjustment is done. If it is embedded in the mp3 decoder, that can be true. If you just use the kmixer... the clipping is already there. I don't know if any player act at the decoding level for volume adjustement, but I don't think so. I may be wrong, though...

Sergio

Edit: fixed a mess that I mad trying to answer to another post.



Would you please invest some of your time and give a try to the solution I suggested?


The FACT is that there is no clipping in the mp3 itself. The clipping is generated at (you guess when?) D_E_C_O_D_I_N_G time.

Sergio

EDIT:
What the heck... I can't post a new answer.. everything goes inside this one... sorry.. don't know what's going on. This is an answer to two different post, thus...

hehe, maybe soon if I do not find a solution

I am unsure of what Moneo suggested really will do the trick.

If I lower the vol (do as Moneo said) and the clipping doesn't occur in the decoded file.
Will the sound wave be prevented to go over that scale, and if so, the wave that normally woudl go over is it cut?

You have alredy been offered a solution but you can't understand it and you are unwilling to even try it. Too bad for you. You'll not be very apreciated in this community with this kind of attitude.

As said, you could use the scale switch with LAME encoder. Values like 0.93 and 0.95 have been recommended, but that is not always enough with the worst examples of over-compresssed recordings. For example, the test track in my screenshot needs something like 0.75 (this depends on the other encoding settings too).

You could analyze the files after encoding with foobar's replay gain, check how much they clip (if at all) and pick the most clipped file, then make a couple of more encodings with estimated scale switch values for finding the highest value that prevents from clipping and finally encode all files again with this value. This is a lot of work.

IMHO, the easy and reliable method is to load the files to MP3gain and let it batch process them to "Maximum no-clip album gain". As I said, this prevents the files from clipping with any decoder. MP3gain does not transcode, it just adjust the volume level. MP3 files are always already altered from the original source and this change can only make them better (IMHO again).

The "clip-prevented" files" can be analyzed for replay gain tags afterwards. Replay gain tags can be used in addition for making the usual playback options available (track or album gain, gain adjustments etc).

come on, I was offered a solution I probably do not agree with

If someone ask you a question and you feel you know the answer but the one that asked it doesn't agree, is that really a solution.
Can and can not be...

Btw I do not think no one will remember me here anyway

I just want to consider other options, easier

Ok I'll be nice and and ask a new easy question.
Does anyone know any vol tool that does the same thing as Volume in SoundForge.
example, you have 10 wav files you have ripped with EAC and all the wav files you want to lower the vol by 0.5-0.6dB (and ofcourse keep the original as before and but make the new wav files with the desired lowered vol.)

Background

With fixed point data, 0dB FS, or “digital full scale” is the largest value you can store - anything bigger will be clipped to that largest possible value.

With floating point data, 0dB is normally defined somewhere near the middle of the possible range. With audio data, typical 32-bit floating point representation gives you over 1000dB range - a lot of space above conventional 0dB FS


So the answer to your question is this: because .wav files from CDs are fixed-point data (16-bit integers), while mp3 files are floating point data.

This particular floating point data is (nearly) impossible to clip. The headroom is amazingly large (at least 100dB more than you need).

mp3s are normally decoded to fixed point data (typically 16-bit integers), though other options are available.


The solution you have been given does work. Try it.


Otherwise, just add the --scale 0.9 switch to lame.

(FAQs and manuals are great things!)

Cheers,
David.

ok ok, I am gonna try but have I understood it correct, you all sayin I do not need to lower the vol at all?

I don't base judgement on "feeling" but on "knowledge", as all the other esteemed members of this community do. You demonstrate the lack of the even the basic knowledge of this matter but where unwilling to accept advices because of your "feelings". That's the problem, man...

ok, you are probably right
But then you probably didn't see the obvious, I needed something easy

"thx for calling me stupid in a nice way"
But then again, stupid is a stupid word, we all lack some kind of knowledge

No, I didn't whish to call you "stupid". You were just unwilling to accept informed advices given on good faith because of your feelings, so maybe "presumptous" would be a more fitting definition, but english is not my mother tongue, so I may be wrong with this term...

Sergio

man, I wasn't offended

we all have 1 consciousness and 1 unconsciousness.
what you maybe really meant you or we may never know

"Imagination is more important than knowledge..." -Albert Einstein (1879-1955)

Sometimes you just go with the feeling

---------

A big thanx to you David (2Bdecided).
Now I know a bit more than before

have to add 1 last but in this topic
but when playing a mp3 file, does it get decoded to wav while playing or not?

Kind of. At some stage, exactly the same type of data (either 16-bit integer or 24-bit integer, stereo PCM samples) has to be generated. Exactly the same data exists in a .wav file as is sent to the sound card driver, but no actual file is generated when you play back mp3s - it's all done "on the fly".

Cheers,
David.

Then I do not see how it can be no clipping in the mp3 file.
I can understand that the mp3 file does not have clipping but the final audio that comes out has clipping
I hope I understood everything correctly, or mayeb not?

Let's say you're a librarian at a big old-fashioned library. You got a book, an on one page there's a circle, filling up the whole page. There are lots of pages like this.

You want to store the book on microfilm. You load the pages in the machine (whatever it's called) and you take a photo of the page with the circle. The film now has an image of a circle.

You want to print some pages from the microfilm. You need to adjust the zoom/magnification.

Oops, now you notice that one of the pages from the book had an incomplete circle; it was cropped at the edges. There's nothing you can do, because the microfilm copy is wrong too.

Oops, you now notice that you made a mistake when you printed the microfilms. You set the zoom too high, and now only part of the circles are on paper. You can adjust the zoom and print again.

---

Easy to understand?

Exactly. As already said, clipping can occour at decoding time of an mp3 stream. mp3gain (and other tools) can do things in order to prevent such clipping.

Really, try mp3gain, (or foobar and its RG featurs). It's SO easy... you just load a bunch of files in it and let it do his tricks... nothing complicated. You, by now have probably anderstood that the difference bewteen the two is that mp3gain (losslessy) alter the mp3 stream so that clipping will nevr occur again with that stream independently of the player used, while foobar with RG embed information so that only it (and few others players) will not clip with that stream.

About your considerations about the importance of feeling and instinct in life, I concur. But if 1, 2, 3, ..., n persons give you an informed advice, on your request, investing their time to help you and you discard their advices "a priori", how do you call that?

Cheers!

Sergio

kjoonlee, you are saying the original has clipping too, but it doesn't if you measure with the limits that wav has.

smz, that choice was for you to make

I have myself been a fool to answer sometimes but ofcourse totally different subject and forum.

-----------

I remember a long time ago when I tried mp3gain and the limit that was recommended was 89dB


and ok, I get the point, mp3 has a lot higher limit than wav.

but how do I do this while ripping or converting, --scale .89 or foobar?

Edit:
I've tried with scale, 89dB is ok
Am I right by saying it just limits the sound wave to go over 89dB and doesn't cut anything?

No. --scale 0.89 doesn't have nothing to do with 89 dB. It convert each and every sample to floating point and then multiply it by 0.89 before feeding it to the encoder.

Notice how I said only some pages of the book were cropped. Only some CDs have clipping.

If you compress CDs, you're bound to get some new clipping sometimes, but the new clipping shouldn't be very serious. I'm just saying you can adjust the volume of the MP3 decoder to prevent clipping.

Please have a look at this picture:



The blue curve is the sum of the red and green curves which in turn are sines of (slightly) different frequency and phase. The blue curve represents the signal stored in your original WAV. For sake of argument, the legal -- non clipping -- range of your wav is from -1 to +1. As you can see, the blue curve is doing fine.

NB the following is grossly simplified:
The mp3 sees that the blue curve consists of two sines of different frequency. It decides that the frequency of the red sine is inaudible and thus removes it. The green curve is untouched. Upon decoding you are only left with the green sine which is obviously out of bounds.

So the clipping only happens when you want to decode (there is no clipping inside the mp3 file). You have to look at what is stored in the mp3 and if the resulting decoded signal exceeds the legal range and by how much. Once you have the maximum amplitude you can use it to calculate the necessary attenuation you have to apply to your mp3 before decoding to wav. The attenuated file will allways be quieter than the original.

The tools to do this have allready been mentioned.

You can of course also attenuate the input signal before encoding, but by how much can never be told beforehand. You also lose some dynamic range (which can be mostly prevented if you use the preferred replaygain/mp3gain approach and a dithering decoder).

Ok, thanx Gecko!

Now I have tried Mp3Gain, and when decoding the mp3 file to wav I see in Sound Forge that the max output is on -8.4dB :S
So my conclusion is that it lowers too much

When lowering in Sound Forge by 0.6-1.0dB on the original wav file it usually does not give clipping on the decode file from the mp3 file.

Thanx Moneo, I'll use the scale 9 command in lame.

And thanx everyone else that helped

I'm thinking that you haven't configured mp3gain correctly. With max no-clip gain applied, the peak level should be no less than -1.5 dB for the track with the highest peak in album mode or every track in track mode.

It's taking a while, but you're getting there.

hehe, probably

Do you mean in Advanced Options, Enable "Maximizing" Features ?

The scale command is ok for me

Yes, maximizing features!

The scale command will require more work. Why do you keep fighting this?

It's quite quick in mp3gain, isn't it? I think there's even a guide to it somewhere, but basically...

1. Back up all your mp3s
2. Options>Advanced> tick "enabled maximising features" (you've already found this)
3. Add all your mp3s to the list
4. Modify Gain>Apply Max no-clip Gain for Album
5. wait - that's it!

You don't have to re-rip anything! Note: MP3Gain automatically assumes that mp3s in different folders are in different "albums".

The problem you'll hit with using --scale to avoid clipping is that however low you set it (unless it's stupidly and audibly low), you'll probably find an album one day that still clips with whatever setting you've chosen, and have to lower it again.


Remind me again: can you actually hear the clipping, or do you just not like knowing it's there?!

I ask because many people (myself included) usually can't hear it - at least when the clipping had been introduced by mp3 encoding of the kind of content which typically causes this problem. Some people can hear it, and on some content I think most people would hear it, but generally it's not a seriously audible problem. (whereas clipping anywhere else is often nasty, and easily audible).

Cheers,
David.

I have found that the clipped peaks that are caused by decoder clipping are often very short, usually less than a millisecond or so. I guess that is a too short duration for a human brain to be detectable. The actual duration of the clipped peaks can be easily seen with any decent wave editor when the display is enlarged to the maximum possible size.

Please explain what do you mean by "lowering the vol"...

mp3gain does exactly that: it lowers (or increase) "the vol" of your mp3 files at your heart content, in setps of 1.5 dB, either by a constant value that you indicate or by a computed value to target either the max non clipping volume, a predefined average album "volume" (typically 89 dB, but can be changed), or a "track" value, so that all the tracks sounds at about the same volume (this last option not adviced, IMHO, unless for DJ purpose, maybe).

Beware that if I'm not misunderstood the "maximize" feature apply different gain values to different tracks in order to maximize EACH track. So you loose the inherent volume differences between the various tracks of an album. Read the docs.. they talk about that...

Sergio

It has an album mode:

This keeps the relative track volume levels inside each album in correct proportion.

Yeap, you're right Alex B! Sorry, but I hadn't mp3gain available and was trying to remember!

Tnxs!

Sergio

Edit: and I was half-blind, as I didn't noticed what 2Bdecided had already written!

Edit2: But I insist that "Album gain" is the way to go, eventually with the target value modified for the few albums that do clip at the "standard" target of 89dB, so that you don't have to adjust your player volume when switching from an album to another one. But this is just a matter of taste...

Point is: MP3's dynamic range is larger than the dynamic range of your WAV files (more room below and more room above). MP3s can represent sample values above and below the limited WAV range (having fixed-point 16 bits in mind).

Since WAV->MP3 is a lossy process the signal is altered but it doesn't involve clipping because MP3 has a larger dynamic range and can represent the signal. But MP3->WAV might force the decoder to clip samples because WAV can only carry a limited range of sample values.

Luckily there's something like Replaygain and MP3gain around. They affect the volume before the mapping to a limited set of possible sample values.

EDIT: Darn! I was replying to the last post on the first page. I didn't notice there was a second page with lots of other replies. ;-)
@Gecko: Nice graphical example!

Coz, I want it to be as good as it can be

But I'll probably start using lossless as soon as I buy a flash mp3 player that supports FLAC.
iAudio U3 has it but 2GB ain't much .........




thanx


I am sure now, I will just use the scale


------

About the clipping, ofcourse I do not hear it.
But as I said b4, the better the better

You still don't (or don't want to) understand. This will not be "as good as it can be". You are going to loos dynamic range. Not with mp3gain (or ReplayGain).

The only drawback I see is: You can't be sure about whether 0.95 will do the trick or not.
But I think I know what you mean by "loos dynamic range". This would be true if LAME quantizes the samples back to 16 bit integers after applying the gain. It does not. LAME imports anything to floating-point and then applies the gain (possible round-off errors when dealing with floating point are negligible in this case).